putative role for cytokines, adhesion molecules and iNOS in brain response to ischemia.

Brain Pathology 10:95-112(2000)Figure 1.Responses of normal cerebral microvessels (A) to focal ischemia, as with middle cerebral artery occlusion (MCAO), include increases in permeability of the endothelial cell component of the blood brain barrier (B), adhesion of poly-morphonuclear (PMN) leukocytes and platelets to endothelial cell receptors expressed in sequence (C) (see text), and loss of integrin-matrix attachments of endothelial cells and astrocyte end-feet which accompany loss of the basal lamina (D).Figure 2.Effect of MCAO and reperfusion on integrin ␣1␤1 expression (hours).sion receptors are expressed in cerebral microvessels.These lead to occlusion of the ischemic microvascula-ture by activated PMN leukocytes and platelets (99).Rapid decrease in expression of integrin receptors on endothelial cells and astrocyte end-feet,together with changes in matrix structure,are initiated within minutes of MCAO in the non-human primate striatum. These changes in receptor expression coincide with the earliest evidence of neuron injury (113),matching both the tem-poral course and extent of non-vascular cell injury.These changes are not altered by reperfusion of the occluded MCA even within hours of the initial occlu-sion. Given the heterogeneous nature of the microvascu-lar response,it is likely that injuries caused,in part,by inflammatory cells lead to confluent destruction of cere-bral tissue in the ischemic territory. However,the early microvascular responses to ischemia,including leuko-cyte adhesion and transmigration,together with struc-tural damage caused by the rapid loss of integrin-matrix adhesion,may impact neuron injury.Figure 3.Changes in expression of the integrins ␣1␤1, ␣6␤4,and ␣V ␤3in relation to the matrix ligand laminin-1 by 2 hours MCAO.after induction of ischemia,the inhibitors were adminis-tered starting 12-24 hours after MCA occlusion,at the time when iNOS was present. It was found that iNOS inhibition reduced infarct volume by 30-40% (72,74, 115). Importantly,the reduction in histological damage was associated with an improvement of the neurological deficits produced by the infarct (103). The effect could not be attributed to changes in body temperature,plas-ma glucose,hematocrit,arterial pressure or blood gases. Furthermore,aminoguanidine did not influence cerebral blood flow (CBF),suggesting that the protection is not related to the preservation of post-ischemic blood flow (74).To rule out the possibility that the protection exerted by iNOS inhibition was related to non-specific effects of the inhibitors,mice lacking the iNOS gene (90) were used in brain following MCA occlusion (73). It was found that iNOS null mice have smaller infarcts (-30%) and better neurological outcome than wild-type litter-mates (73). The reduction in infarct volume was more marked in homozygous than in heterozygous iNOS mice (164). This observation is consistent with a gene-dosing effect of iNOS deletion. The protection could not result from cerebrovascular effects of iNOS deletion, because the reduction in CBF produced by MCA occlu-sion did not differ between iNOS null mice and controls (73). Furthermore,the reduction in infarct volume could not be attributed to effects on the cellular reaction that occurs after ischemia,because the degree of neu-trophilic infiltration and astrocytic activation was com-parable in iNOS null mice and controls (73).Interestingly,the magnitude of the protection exerted by deletion of the iNOS gene is dependent on the age of the mice. Thus,the reduction in infarct volume was greater in 1-2 month-old mice than in 6 month-old mice (102).iNOS as a therapeutic target for strokeThe findings reviewed above,collectively,provide strong evidence in favor of a major role of NO produced by iNOS in the mechanisms of the evolution of cerebral ischemic injury. Furthermore,the fact that iNOS is expressed also in the human brain after ischemia, strengthens the argument that iNOS is a valuable thera-peutic target in human stroke. The extended therapeutic window of iNOS inhibitors (12-24 hours) would permit to treat stroke patients that do not qualify for treatment with modalities,such as thrombolysis or glutamate receptor inhibition,that are effective only in the early stages of the damage (37). Therefore,inhibition of iNOS expression or activity would be a valuable therapeutic strategy to selectively target the delayed phase of the damage.Cyclooxygenase gene expression and inflammation Another gene that is expressed during inflammation is cyclooxygenase-2 (COX-2). Cyclooxygenases are rate limiting enzymes in the synthesis of prostaglandins and thromboxanes. Two isoforms of COX have been described:COX-1 and COX-2. COX-1 is expressed in many cells and is thought to play a role in platelet aggre-gation,gastric secretion,and renal function (136). COX-2 is constitutively expressed in excitatory neurons, wherein it is localized to dendritic spines (79,158). In many organs,COX-2 expression is upregulated by a wide variety of stimuli,such as inflammatory mediators and mitogens (137). In models of inflammation,COX-2 reaction products are believed to be destructive and con-tribute to cytotoxicity possibly due to production of reactive oxygen species and toxic prostanoids (129). COX-2 and cerebral ischemiaFollowing cerebral ischemia,COX-2 mRNA and protein are upregulated peaking 12-24 hours after ischemia (95,109,110,119) (Figure 4). COX-2 is expressed in neurons and vascular cells located at the border of the ischemic territory (95,110) (Table 3). In neurons,COX-2 is expressed in cells that exhibit ischemic changes,as well as in neurons that appear structurally normal (110). Recently,COX-2 has been found to be expressed also in the human brain after ischemic stroke (69).The role of COX-2 in the mechanisms of cerebral ischemia has not been completely elucidated. Initial data suggest that the relatively selective COX-2 inhibitor NS-398 reduces infarct volume by 20-30% in a model of focal ischemia (110). Furthermore,COX-2 inhibition also reduces neuronal damage in a model of global cerebral ischemia (104). These observations raise the possibility that COX-2 reaction products contribute to the evolution of ischemic damage. The fact that delayed administration of NS-398 (6 hours after MCA occlu-sion) reduces the damage supports the notion that COX-2 is involved in the late stages of ischemic injury. However,evidence that NS-398 acts exclusively on COX-2 activity is lacking. NS-398,like other COX inhibitors,may also have effects on gene transcription which may play a role in its protective effect (54,80).Another line of evidence,suggestive of a pathogenic role of COX-2 in the mechanisms of cerebral ischemia, is provided by studies in which the interaction between NO produced by iNOS and COX-2 was investigated.Following cerebral ischemia,iNOS and COX-2 are expressed with a similar time-course and in cells that are close to each other (109) (Figure 4,Table 3). The spatial and temporal proximity of iNOS and COX-2 suggests that NO produced by iNOS could activate COX-2 and enhance the toxic output of the enzyme (126). This pos-sibility is supported by studies demonstrating that selec-tive inhibition of iNOS reduces COX-2 reaction prod-ucts in the post-ischemic brain (109). Furthermore, COX-2 reaction products are reduced in iNOS null mice (109),which do not produce iNOS-derive NO after ischemia. These data,collectively,suggest that NO pro-duced by iNOS may “drive”COX-2 activity in the post-ischemic brain and increase COX-2 reaction products. COX-2 as a therapeutic target for strokeWhile the interaction between iNOS and COX-2 pro-vides additional evidence that COX-2 activity may be deleterious to the ischemic brain,the role of COX-2 in ischemic brain injury is far from clear. It has long been known that COX reaction products contribute to the reg-ulation of the cerebral circulation (118). COX-2 inhibi-tion might have effects on cerebrovascular regulation that could alter the outcome of cerebral ischemia. These effects need to be characterized and their contribution to ischemic injury remains to be defined. Furthermore, COX-2 is expressed in excitatory neurons,wherein it is likely to play a role in synaptic transmission and plas-ticity (79). Therefore,COX-2 inhibition may have effects on regenerative and processes involved in func-tional recovery after stroke. Further studies addressing these issues are required to clarify in full the role of COX-2 in ischemic brain injury.Molecular mechanisms of expression of iNOS and COX-2The molecular mechanisms of iNOS and COX-2 expression after cerebral ischemia are not fully under-stood. In a variety of cell systems,cytokines are known to induce expression of both iNOS and COX-2. Furthermore,hypoxia has been reported to elicit iNOS expression via activation of the hypoxia inducible fac-tor-1 (94). Transcription factors that are involved in the expression of inflammation related genes include NF␬B and interferon regulatory factor-1 (5,108). These tran-scription factors are induced after ischemia (20,70) and they are likely to contribute to the expression of iNOS and COX2,as well as other genes expressed after cere-bral ischemia. Inflammatory cytokine in brain ischemia and trauma Direct trauma,deprivation of oxygen and nutrients (ischemia),neurotoxicity,viral infection or immunolog-ical challenge produce a well-defined response of “glio-sis”(117). The activation,proliferation and hypertrophy of cells derived from the mononuclear phagocytic sys-tem (e.g.,macrophages and microglia) are the hallmark of this reaction. Originally,this response was thought to mediate repair,restoration of blood supply,re-establish-ing the integrity of the blood brain barrier and promot-ing general homeostasis at the site of injury (111,112). Since cytokines activate glial cells (39) which then pro-duce cytokines (128),a close relationship appears to exist between inflammation,cytokine production and gliosis. Indeed,gliosis can be induced by TNF-␣,IL-1␤and interferon ␣(IFN␣) (6).Several cell types within the brain are able to secrete cytokines,including microglia,astrocytes,endothelial cells and neurons; in addition,there is also evidence to support the involvement of peripherally derived cytokines in brain inflammation. Peripherally derived mononuclear phagocytes,T-lymphocytes,natural killer (NK) cells and PMN’s which produce and secrete cytokines,can all contribute to CNS inflammation and gliosis. In support of this,irradiation of the bone marrow or treatment in vivo with colchicine,attenuate gliosis, wound repair,neovascularization and generalized inflammation (47).The inflammatory response to brain injury has been studied systematically following focal stroke by several investigators. The early accumulation of neutrophils in ischemic brain damage has been clearly demonstrated based upon histopathological (25,43,56),biochemical (9),and 111ln-labeled leukocyte studies (38). Ameboid microglia,a form of activated microglia,can be identi-fied within 2 hours of ischemia. Unlike normal brain microvessels that are clear of inflammatory cell brain microvessels from ischemic zones are filled with leuko-cytes and a significant zone of edema surrounds them. Many of the leukocytes,primarily neutrophils,found in vessels within ischemic tissue are adherent to the endothelium,a situation not normally observed in intact brain microvessels. Some of these neutrophils migrate outside the vascular walls into the focal ischemic cortex. Brain injury is associated with the expression of inflam-matory mediators,e.g.,inflammatory cytokines (IL-1 and TNF-␣) and chemokines,(IL-8 for neutrophils, MCP-1,RANTES,IP-10) while up-regulation of adhe-sion receptors (ICAM-1,selectins) support leukocyte adherence to the endothelium (41). TNF-␣and IL-1␤pre-dispose or “prime”endothelium for cellular adher-ence. Additionally,adhesion molecules such asCD11/CD18 integrins are also thought to be pivotal inthis inflammatory process. The importance of leukocyteinfiltration in the pathogenesis of brain injury has beenreviewed previously (41). This inflammatory reactionnot only contributes to lipid-membrane peroxidation,but also exacerbates the degree of tissue injury due tothe rheologic effects of “sticky”leukocytes in the bloodvessels (i.e.,an interference with normal microvascularperfusion due to vascular plugging in an already com-promized ischemic tissue bed),and also due to therelease of cytotoxic products from these activated leuko-cytes (i.e.,by generation and release of oxygen radicalsand cytotoxic products that are cytodestructive to thealready compromised tissue; 82). The exact nature ofthe signaling mechanisms in brain inflammation stillremains to be elucidated but undoubtedly involves TNF-␣and IL-1␤,chemotactic cytokines (e.g.,chemokines such as IL-8) as well as the expression of adhesion mol-ecules and proteinases that together promote bothrecruitment of adherent leukocytes and infiltration,andenhanced permeability of brain endothelium. For exam-ple,for a neutrophil to adhere to the endothelium andthen migrate unidirectionally into the tissue,adhesionmolecules have to be up-regulated. In order to upregu-late adhesion molecule expression,injury must up-regu-late specific cytokines,such as TNF-␣and IL-1␤. This has to happen very rapidly,and proteins also have to be rapidly translated. In addition,when those cytokines are expressed and translated,they have to up-regulate adhe-sion molecules and produce chemokines that will induce chemoattraction to drive neutrophils into the tissue.Roles of TNF-␣in Traumatic Brain Trauma and Stroke TNF-␣is a pleotrophic cytokine released by many cell types upon diverse stimulation. TNF-␣exerts a diverse array of biological activities including secretions of acute phase proteins and vascular permeability. TNF-␣and its receptors are present in the CNS (135). In addi-tion,several clinical studies have shown a distinct rela-tionship between elevated levels of cytokines,including TNF-␣,neurodegenerative disorders,and brain injury.Elevated TNF-␣has been repeatedly demonstrated in various experimental models of brain injury. Systemic kainic acid administration induces within 2-4 hours TNF-␣mRNA levels in cerebral cortex,hippocampus and hypothalamus. Systemic or intracerebroventricular administration of lipopdysaccharide endotoxin (LPS) has also been shown to increase brain TNF-␣levels as determined by bioassay (134). In a model of closed head injury,Shohami et al. (133) reported an early increase in TNF-␣peptide at the site of the focal insult. Also,in rat traumatic head injury,TNF-␣mRNA and protein levels are rapidly elevated (40). Furthermore,in mice chal-lenged with particles of charcoal injected into the hip-pocampus,an increase in striatal levels of TNF-␣mRNA was observed (133). Elevated serum TNF-␣was also observed following severe head injury in man (49).Elevated expression of TNF-␣mRNA and protein occurs shortly (1-3 hours) following middle cerebral artery occlusion (MCAO) in rats (87,154). In ischemic cortex,TNF-␣mRNA levels are elevated as early as 1 hr post-occlusion (i.e.,prior to significant influx PMN) peaked at 12 hours and persisted for about 5 days. The early expression of TNF-␣mRNA preceding leukocyte infiltration suggests that TNF-␣may be involved in this response. Double-labeling immunofluorescence studies localized the de novo synthesized TNF-␣to neurons but not astroglia. At 5 days following the ischemic insult, neuronally-associated TNF-␣was diminished,and TNF-␣immunoreactivity was localized in the inflam-matory cells. The significance of TNF-␣expression in the brain was studied by microinjection of TNF-␣into the rat cortex; TNF-␣induced leukocyte adhesion to the capillary endothelium,but no evidence for neurotoxici-ty at the site of injection was found. Buttini et al. (18) identified a rapid upregulation of TNF-␣mRNA and protein in activated microglia and macrophages follow-ing focal stroke,again suggesting that TNF-␣is part of an intrinsic inflammatory reaction of the brain following ischemia. TNF-␣may exert a primary effect on microvascular inflammatory response as reflected by TNF-␣-induced neutrophil adhesion to brain capillary endothelium (87). Furthermore,intracerebroventricular injection of TNF-␣24 hr prior to MCAO exacerbates the ischemia induced tissue injury (8). This effect was reversed by ventricular administration of anti-TNF-␣mAb in the contralateral ventricle. Further evidence for the involvement of TNF-␣in stroke-induced injury is supported by findings that spontaneously hypertensive rats that are stroke prone have higher levels of TNF-␣production in the brain as compared with normotensive rats (134). These data suggest that TNF-␣may prime the brain for subsequent damage by activating capillary endothelium to a pro-adhesive state.IL-1␤in brain trauma and strokeIL-1␤is produced in the CNS by various cellular ele-ments including microglia,astrocytes,neurons and endothelium (124). Like TNF-␣,IL-1␤has many pro-inflammatory properties ,and receptors for this cytokine have been demonstrated in the CNS. Increase in IL-1␤mRNA expression has been shown to occur following several types of injury to the brain including kainate excitotoxicity (96) and LPS (19). Furthermore,mechan-ical damage following implantation of a microdialysis probe has been shown to induce expression of IL-1␤. Following fluid percussion brain trauma in the rat,a rapid increase in IL-1␤mRNA expression has been reported. Microglial IL-1␣expression has been observed in human head injury. IL-␤mRNA expression has been shown to increase following transient brain ischemia in the rat (97). The exacerbation of ischemic brain injury due to exogenous IL-1␤administered into the brain has been observed (159). A rapid (3-6 hr post ischemia) increase in IL-1␤mRNA following MCAO peaked at 12 hours but returned to basal values at 5 days (88,154). Early IL-1␤expression following focal stroke has also been demonstrated using in situ hybridization. The recent development of tools such as specific anti-bodies to rat IL-1␤has permitted the identification (by immunohistochemistry) of IL-1␤peptide in cerebral vessels,microglia and macrophages following focal stroke.Interleukin-1 receptor antagonist (IL-1ra),a 23-25 kDa-glycosylated protein,is a naturally occurring inhibitor of IL-1 activity that competes with IL-1 for occupancy of IL-1RI without inducing a signal of its own. IL-1ra is produced by many different cellular sources including monocytes/macrophages,endothelial cells,fibroblasts,neurons and glial cells. The expression of IL-1ra and IL-1R mRNA following focal stroke were also reported (150). The level of IL-1ra mRNA was markedly increased in the ischemic cortex at 6 hours, then reached a significantly elevated level from 12 hours to 5 days following MCAO. The presence of IL-1ra in the normal brain and the upregulation of IL-1ra mRNA after ischemic injury suggest that IL-1ra may serve as a defense system to attenuate the IL-1-mediated brain injury. It is interesting to observe that the temporal induction profile of IL-1ra following MCAO virtually parallels that of IL-1␤as demonstrated previously (88), except that IL-1ra mRNA exhibited prolonged elevation beyond that of IL-1␤. Thus,the balance between the levels of IL-1␤and IL-1ra expressed post ischemia may be more critical to the degree of tissue injury than IL-1 levels per se.The mediators responsible for IL-1ra induction after focal stroke are not known. However,previous studies indicate that some cytokines such as IL-1,TNF,IL-6 and TGF␤are inducers of IL-1ra. Ischemia-induced expression of IL-1ra mRNA could originate from mono-cytes/macrophages,endothelial cells,fibroblasts,neu-rons and glial cells as observed previously under normal conditions. The same cellular sources may be responsi-ble for IL-1␤and IL-1ra production based upon the close temporal,and perhaps functional coupling of these two genes after focal stroke.Anti-Leukocyte Strategies for Neuroprotection Matsuo et al. (92) have used the RP-3 monoclonal antibody that selectively depletes leukocytes in the rat (about 90-95%) and reported a dramatic reduction in both neutrophil accumulation in focal ischemic brain tissue and infarct size (decreased by 45-50%). However, some controversy exists (62).Anti-adhesion molecule antagonists for neuropro-tectionAnother attractive approach is the inhibition of endothelial interactions with the leukocyte. Chen et al.(21) treated MCAO rats intravenously with an antibody against MAC-1,the leukocyte counterpart of ICAM-1 binding and demonstrated reduction in infarct size by 45-50% in a rat transient MCAO model. Zhang et al. (162) used the i.v. administration of an anti-ICAM-1 antibody to demonstrate a 40% reduction of infarct size in a similar model. Blocking adhesion molecules can also reduce apoptosis induced by focal ischemia (23). Other studies verified these effects but also illustrated that these antibodies could not reduce infarct size when the ischemia was permanent (15,24,26,77,161). However,the strategy may work if both leukocyte and endothelial adhesion proteins are blocked in permanent focal stroke. The combination of t-PA and anti-CD 18 provides significantly improved outcome,and may increase the therapeutic time window in stroke (163). In a rabbit embolic model of stroke,anti-ICAM-1 antibody was shown to increase the amount of clot necessary to produce permanent damage (15). In addition,in a baboon model of transient focal ischemia anti-CD 18 monoclonal anti-body administered 25 min prior to reperfusion led to increase in reflow in microvessels of various sizes (99). However,in contrast to the demon-strated anti-ischemic effect of anti-adhesion molecules in animal models,the recent failure of the murine anti-ICAM mAb (enlimomab) in human stroke (31) and its ability to activate human neutrophils (148) demonstrate the difficulties in extrapolating encouraging data derived from animal models to clinical reality. Neuroprotection by cytokine inhibition supports anti-inflammatory approachAn alternate possibility to modulate inflammation isto aim directly for cytokine and chemokine suppressive agents. While proof for a role of TNF-␣and IL-1␤in ischemia brain damage has not been definitively estab-lished,the availability of selective and potent antago-nists of cytokine production may aid in reaching this goal. Many studies have demonstrated the protective effects of IL-1ra in brain injury. Thus,intracerebroven-tricular administration of recombinant IL-1ra produced a marked reduction in brain damage induced by focal stroke (89,120,125),or brain hypoxia (91). This neu-ronal protective effect of IL-1ra in focal stroke was fur-ther supported by a recent study using an adenoviral vector that over-expressed IL-1ra in the brain (12). The excess of IL-1ra significantly reduced infarct size fol-lowing focal stroke. While such modes of IL-1ra deliv-ery are impractical in clinical terms,the studies point out a potential therapeutic remedy if delivery of IL-1ra can be achieved in a timely fashion. In addition,IL-1ra expression increases following ischemic precondition-ing in a manner that parallels the development of brain ischemic tolerance (11). Of interest is data showing that peripheral administration of IL-1ra reduces brain injury (120),suggesting a potential use of IL-1ra as a neuro-protective agent in human stroke and/or neurotrauma. Likewise,several studies have shown that blocking TNF-␣results in improved outcome in brain trauma and stroke. Pentoxifyline (a methylxanthine that reduces TNF-␣production at the transcriptional level) or soluble TNF receptor I (which acts by competing with TNF-␣at the receptor) improves neurological outcome,reduces the disruption of the blood brain barrier and protects hippocampal cells from delayed cell death following closed head injury in the rat (131). In rat focal ischemia, an anti-TNF-␣monoclonal antibody (mAb) and the sol-uble TNF receptor I were neuroprotective (8). In the lat-ter studies,TNF-␣was blocked by repeated i.c.v. administrations before and during focal stroke which significantly reduced infarct size. In murine focal stroke, topical application of soluble TNF receptor I on the brain surface significantly reduced ischemic brain injury (106,107). In addition,in another study evaluating TNF blockade on focal stroke in hypertensive rats,soluble TNF receptor I administered i.v. pre- or post-MCAO significantly reduced the impairment in ischemic cortex microvascular perfusion and the degree of cortical infarction,strongly suggesting an inflammatory/vascu-lar mechanism for TNF-␣in focal stroke (30).The detrimental effects of TNF-␣and its role as a mediator of focal ischemia may involve several mecha-nisms. For example,TNF-␣increases blood brain barri-er permeability and produces pial artery constriction that can contribute to focal ischemic brain injury; fur-thermore,a direct toxic effect of TNF-␣on the capillar-ies were also noted (48). Furthermore,by stimulating the production of matrix-degradating metalloproteinase (gelatinase B) (122,123),TNF-␣may further exacer-bate capillary integrity. TNF-␣also causes damage to myelin and oligodendrocytes (121),and increases astro-cytic proliferation thus potentially contributing to demyelination and reactive gliosis . In addition,TNF-␣activates the endothelium for leukocyte adherence and procoagulation activity (i.e.,increased tissue factor,von Willebrand factor and platelet activating factor) that can exacerbate ischemic damage. Indeed,increased TNF-␣in the brain and blood in response to lipopolysaccharide appears to contribute to increased stroke sensitivity/risk in hypertensive rats (10,41,134). TNF-␣activates neu-trophils increases leukocyte-endothelial cell adhesion molecule expression,leukocyte adherence to blood ves-sels,and subsequent infiltration into the brain (41).Interference with either IL-1 or TNF-␣has now been shown repeatedly to result in reduced deficits in focal stroke and head trauma models. The evaluation of addi-tional potent and specific anti-cytokine therapeutics in proper models of brain injury is clearly warranted. Much evidence has accumulated that indicates TNF-␣production is regulated at both transcriptional and trans-lational levels (144). Thus,TNF-␣mRNA synthesis inhibitors such as rolipram (130),a phosphodiesterase IV inhibitor,could be of benefit in the treatment of brain inflammation. Other novel classes of drugs include highly specific protein kinase C (PKC) inhibitors such as calphostin C (81),which has been shown to potently inhibit LPS-stimulated TNF-␣production from human monocytes in vitro as well as LPS and viral stimulated TNF-␣production in astrocytic cell lines (85). However, due to the ubiquitous nature of PKC,non selective inhibitors may result in toxic consequences. The utility of blocking inflammatory cytokines in conjunction with thrombolysis using tPA has been discussed recently (44).Cytokines and brain pre-conditioningWhen exposed to single or repetitive episodes of sub-lethal ischemic stress,brain cells acquire resistance to subsequent,otherwise lethal ischemic insults (for review see 22). As with other types of stress,ischemic stress causes a number of biochemical changes in the cell which trigger activation of multiple signaling path-ways. In turn this leads to expression of new genes and/or down regulation of currently active genes. Some of these changes are beneficial such as redistribution ofenergy,activation of alternative metabolic pathways,production of antioxidants,and activation of DNA repair mechanisms. Others,like release of cytokines,upregulation of adhesion molecules,and induction of apoptotic machinery,exacerbate brain injury. In response to the stressful situation,another wave of bio-chemical and genetic changes occurs which attempts to resist the stress and to return the brain to its original con-dition. Thus,the same stress stimuli could cause both cell protection and cell death and also could elicit feed back and/or feed forward reactions that would quench either of these responses. This quenching reaction con-sists of at least of two mechanisms:either cytoprotective genes are activated to neutralize the effect of cytotoxic gene products (e.g. anti-inflammatory cytokines vs.proflammatory cytokines or anti-apoptotic genes vs.pro-apoptotic genes) or expression of harmful genes is simply shut off by newly synthesized or activated inhibitors of transcription and/or translation. We suggest that ischemic tolerance occurs because of this “after stress”homeostatic correction. Indeed,development of the tolerant state takes time,usually 24-72 hours (11). It is quite possible that although sublethal stress causes no visible damage,it still initiates anti-stress responses that proceed during the latent period and result in a transient stress-resistant phenotype of brain cells. Within this the-oretical framework,a molecular mediator which triggers ischemic preconditioning could be the same agent that triggers ischemic injury; only in the case of precondi-tioning it is not harmful because the intensity of the effect is held below the threshold of cytotoxicity and/or because its effects are not exacerbated by severe stress environment. The response to such a mediator affects expression of multiple genes in the various types of brain cells.TNF-␣,a cytokine with pleiotropic activity,is a uni-versal agonist for many different types of cells. Thus,whole brain,neurons (57,87),microglia,astrocytes (145),and brain endothelium (14) are not only capable of TNF-␣synthesis in response to stress but also express TNF-␣receptors and amplify this response through paracrine and autocrine mechanisms (135). Further,TNF-␣has been implicated in both,detrimental (3) and neuroprotective (93) actions on brain cells depending on the experimental conditions. Our recent studies demon-strated a mild but significant induction of TNF-␣mRNA in ischemic brain tolerance,suggesting a potential neu-roprotective role of this cytokine (151). The dual func-tion of TNF-␣has been also revealed in vivo (reviewed in 132). Neutralization of TNF-␣by TNF-␣-binding protein had a protective effect against focal ischemiaFigure 5.Cell Death in Neuronal Cultures Subjected to Hypoxia or to Oxygen/Glucose Deprivation (O/GD).Neuronal cultures grown in 24 well plates, either naïve or preconditioned with 20min hypoxia were subjected to 2.5 hours of hypoxia or to 2.5hours of O/GD and percentage of dead cells was measured at 8 hours after reoxygenation by means of ethidium homodimer exclusion fluorescent assay.Dead cells were also measured in control, untreated but sham-washed cultures.Each measure-ment was performed in eight wells and averaged.Each bar rep-resents mean ϮSD of 4 experiments * and ** denotes signifi-cant difference from control and from hypoxia or (O/GD),respectively.Figure 6.The role of TNF-␣in hypoxic preconditioning of neu-rons against hypoxia-induced neuronal injury.Neuronal cul-tures, were preconditioned with 20 min hypoxia in the absence or presence of TNF-␣neutralizing antibody, and then subjected to 2.5 hours of hypoxia.Measurements of the dead cell number were performed at 8 and 24 hours after cell reoxygenation.Each measurement was performed in 8 wells and averaged.Each bar represents mean ϮSD of 4 experiments.* and **denote significant difference from hypoxia-induced injury in naïve and preconditioned cells, respectively.。